Optimize Workbench performance so it's on par with the previous
implementation.
Most of these changes are barely noticeable on powerful GPUs,
but can cause a notable performance improvement on old or low-end
hardware.
* Avoid unnecessary texture copies and draw directly to the viewport
textures.
* Optimize-out depth/stencil reads, using stencil testing instead.
* Avoid using `Texture::clear` and use framebuffer clears instead.
* Avoid framebuffer state changes (always use the same attachments).
* Avoid constant variation of acquired `TextureFromPool`s.
Fix#113010
Pull Request: https://projects.blender.org/blender/blender/pulls/113251
Caused by fb556c75df
The ideal would be to increase the version file, but this would make
the versioning code a bit messy, and it was a recent error
(it shouldn't affect many users).
Since garbage collection of texture pools happens every time
`drw_manager_init` is called, interleaved calls to different
`DRW_draw`/`DRW_render` functions can result in a constant cycle of
creating/releasing textures.
This PR changes texture pools to only release textures that have not
been used for N garbage collection cycles.
I chose 8 as default since I think is high enough to avoid the issue
while it's unlikely to cause any major spike in memory consumption.
Pull Request: https://projects.blender.org/blender/blender/pulls/113024
Shadow Map Ray Tracing is a technique that ray cast against the shadow
depth buffer. The technique is described in "Soft Shadows by
Ray Tracing Multilayer Transparent Shadow Maps".
Note that we only implement the single layer approach since storing
multiple depth is prohibitively expensive.
Pull Request: https://projects.blender.org/blender/blender/pulls/111809
The case when there were no material indices, multiple materials, and
the mesh used hiding wasn't handled after 55970fa367.
This is a simple parallel loop adding up the non-hidden triangles.
Hair and Curves need to call `DRW_curves_update` before
`geometry_steps_fill`, otherwise the copied geometry is just
uninitialized data.
However, doing so triggers an assertion:
> DRW_render_instance_buffer_finish had not been called before drawing.
This PR ports the `DRW_hair/curves_pos_buffer_get` functions to the new
`draw::Manager` API, so it can be called at any arbitrary point.
It also changes `VelocityModule::geometry_map` to use `uint64` keys
instead of `ID` pointers, since the same particle system can be used on
different objects and have multiple geometries.
Notes:
* The new functions are only used for
`VelocityModule::step_object_sync` on image renders. Using them
elsewhere would require modifying the old draw manager to do the
init/update/free setup.
* Only the compute shader version has been ported.
Pull Request: https://projects.blender.org/blender/blender/pulls/112425
Implement a clip distance to irradiance grids, so surfaces farther than
the threshold don't generate surfels.
Align the capture view matrix to the probe rotation.
Pull Request: https://projects.blender.org/blender/blender/pulls/112863
This makes the SSS processing more isolated to avoid
side effects in other passes.
Also this allow more optimization using LDS.
The radiance buffer is created by a setup phase using
direct and indirect lighting.
Splitting this into its own compute pass also fixes
a pass merging issue on Metal.
Pull Request: https://projects.blender.org/blender/blender/pulls/112901
This avoid confusion outside of the shader node GLSL code.
The issue also is that Metal allow float to cast to bool
implicitly but this create a compilation error on
OpenGL.
Scenes that only uses a world probe would didn't work. Raytracing
would look for the closest light probe, didn't find any and shader
would use undefined behavior to sample in the reflectionProbes texture.
In some cases this resulted in correct results when the world was
located on layer 0 with subdivision 0. In any other case this would
sample incorrect incoming light what lead to black reflections.
This PR fixes this by returning the world probe when no nearest light
probe cound be found.
Pull Request: https://projects.blender.org/blender/blender/pulls/112893
This adds a new entry to the split sum LUT to isolate
the effect of the F82 tint.
The application of the tint part is similar to cycles
and uses the same way for precomputing the `b` factor.
Results matches almost perfectly to the extent of the
split sum approximation.
Note that this removes the unused LTC MAG LUT for
EEVEE next to make space for the new table. It can still
be added back if needed.
Pull Request: https://projects.blender.org/blender/blender/pulls/112881
Texture Atomics have been added in Metal 3.1
and enable the original implementations of
shadow update and irradiance cache baking.
However, a fallback solution will be
required for versions under macOS 14.0 utilising
buffer-backed textures instead.
This patch also includes a stub implementation if
building/running on older macOS versions which
provides locally-synchronized texture access in
place of atomics. This enables some effects to be
partially tested, and ensures non-guarded use
of imageAtomic functions does not result
in compilation failure.
Authored by Apple: Michael Parkin-White
Pull Request: https://projects.blender.org/blender/blender/pulls/112866
This keeps the behavior similar to the Disney BRDF, where 0.5
is neutral and lower/higher values respectively decrease/increase
the dielectric specular. But it's more correct in that it's not
an arbitrary scale on Fresnel, but rather adjusting the IOR.
Ref #99447
Ref #112848
Pull Request: https://projects.blender.org/blender/blender/pulls/112552
Enable visual keying of bones that are influenced by an IK constraint.
This wasn't possible before, as the visual keying system only checked
constraints on the bone itself, and not whether the bone was part of an
IK chain.
This commit introduces a new `bPoseChannel::constflag` value
`PCHAN_INFLUENCED_BY_IK` that is set whenever the pose bone is part of
an IK chain.
The `pchan->constflag` field is computed during depsgraph evaluation. If
the depsgraph is active, it is now also written back to the original
pchan, so that it can be used in the "should visual keying be used"
function.
Fixes: #76791 "Different results when keyframing visual transforms and
applying transforms manually on IK constraint". Note that visually
keying does *not* copy the visual pose to the current pose. Furthermore,
when visually keying only part of the IK chain, the result of
re-evaluating the IK constraint (for example by moving the scene forward
and then backward by one frame) may still produce a different result, as
the IK chain now has a different start orientation.
Note that commit explicitly does not cover Spline IK constraints. They
can introduce heavy shear, especially with the default settings, which
cannot be represented by keys on loc/rot/scale.
For historical reference: 876cfc837e
introduces the 'use visual keying' preference option, where Blender
automatically chooses whether or not to use visual keying. This is why
there is a function at all that determines whether to use visual keying
or not.
Rename the `bPoseChannel::flag` `PCHAN_HAS_TARGET` to `PCHAN_HAS_NO_TARGET`
as that is actually the meaning of the flag (in the majority of the code).
Since the flag was so confusingly named, there were some mixups in the
armature overlay drawing code as well, which have been fixed now too.
The Pixelate effect has an antialiasing switch that cleans up
pixels mostly, but still has some inbetween artifacts that are
a little frustrating to deal with. This makes it so that pixels are
completely hard-edged when AA is disabled.
Pull Request: https://projects.blender.org/blender/blender/pulls/112605
Blender 4.0 requires OpenGL 4.3 which always support SSBO's.
Platforms that don't support enough SSBO bind points will be marked
as unsupported.
Users who start Blender on those platforms will be informed via a
dialog. This PR also updates the `--debug-gpu-force-workarounds`
to match our minimum requirements. Note that some bugs are still
there that should be solved in other PRs:
* Workbench only renders the object using a unit matrix this is because
there is a bug in the workaround for shader_draw_parameters
* Navigating with middle mouse button is not working. Unsure what the
cause is, but might be a missing feature check in the OpenGL backend.
Related to #112224
Pull Request: https://projects.blender.org/blender/blender/pulls/112572
This implement the diffuse tracing as simple
screen-space ray trace. This is extremely inefficient
and should only be used for reference purpose.
The real screen space diffuse implementation will
be done in another PR.
Depends on #112507
Pull Request: https://projects.blender.org/blender/blender/pulls/112539
This allow to move the light evaluation up in the deferred pipeline and
avoid coupling two different steps into one. This add more flexibility
in the implementation of deferred lighting and indirect lighting
algorithms.
This splits the reflection probe evaluation to its own
shader since it is quite different now.
Pull Request: https://projects.blender.org/blender/blender/pulls/112507
This moves the pre-computation offline and store the pre-computed
table in the binary. The pre-computed tables are quite small and are
not a concern with respect to binary size increase.
This rewrites the precomputation to use manually fitted
approximations for both burley and random walk.
The approximations fix a discrepancy between cycles and EEVEE
SSS translucency look. The absolute maximum error is below 2%.
I believe better results could be achieved with automatic fitting
tools.
Note that Cycles Burley translucency profile has some issues as it
does not give a smooth profile. The profile is biased near the end
of the lower radii. For this reason, the fit was done on a white
diffuse with (1,1,1) radii which does not exhibit this artifact.
Note that while this adds the profile for random walk, it isn't
currently used because the profile type is not yet passed down
the deferred path.
The fitting data can be found attached to this PR.
Pull Request: https://projects.blender.org/blender/blender/pulls/112512
This allow to move the light evaluation up in
the deferred pipeline and avoid coupling two
different steps into one. This add more
flexibility in the implementation of
deferred lighting and indirect lighting
algorithms.
This splits the reflection probe evaluation
to its own shader since it is quite different
now.
